When Algorithms Offend: Unpacking the Causes and Cures for User Aversion in AI Recommender Systems

This study examines the causes and mitigation of user aversion to AI-generated recommendations. It identifies perceived intrusiveness, information bubbles, and content overload as key drivers, with intrusiveness being the most influential due to privacy and autonomy concerns. Using survey data and SEM, results show these factors increase cognitive and emotional burden. Mitigation strategies reveal that optional pause features most effectively reduce aversion by enhancing user control, while social recommendations provide moderate relief. The study extends Reactance Theory and Cognitive Load Theory to AI contexts and offers design and policy insights to improve user trust and autonomy.